Let's made a PC-base metal detector with usb interface !!!

[Aziz]

Doubling the transmit power has increase of approx. 12% of depth distance. I am not sure, whether I have calculated this correct.

Doubling the Power: factor 2
Depth increase (percent): pow(2, (1/6)) = 1.12246 = ~12% increase

pow(2, 1/6) = sixth square root of 2

Other calculation (reducing power):
Half power: factor 0.5
Depth (percent): pow(0.5, (1/6)) = 0.8909 (~11% distance loss)

Please correct me, when I am wrong!

Aziz

[JC1]

64 times

Hi Aziz,

or said another way you would need to

increase xmit power by 64 times to double

the detection depth, changing nothing else.

or you could increase the receive gain by say 10,

and xmit power by 6.4 and this would also work.

assuming you can increase receive gain by 10 or more

without noise becoming a problem.

(this is total gain through the entire chain of amplification)

You can also make the coil larger for larger targets.

By combining these three you can do well, but it is

always a struggle past a certain point.

have fun.

I have.

[Aziz]

Hi guys,

using center-taped RX coil can increase the dynamic range to from 100 mV rms to 200 mV rms. This will give a better dynamic range.

We have two input channels (left/right). The center point is connected to ground. Both other coil connections are connected to left and right channel. The signal in the laptop has to be subtracted:
signal = left-right or
signal = right-left

I will make some experiments using center-taped RX and TX coil to break new detection distance records. Now the coil is drying...
I know now, how the software should handle all different configurations. It has to be more flexible.

Aziz

[Aziz]

Why need separate sound card because of two frequencies? Most sound cards have at least stereo don't they? One for TX signal, one for headset. TX signal can be sum of two frequencies. Make coil with two resonant frequencies. No?

Hi Simon,

it is difficult to implement this feature due to fixed RX-resonant circuit (RX LC part). So it is better to use one frequency (TX) and implementing 3 individual lock-in amps:
- one on the resonant frequency fr
- one on fr-BW
- one on fr+BW

BW is the bandwidth around the resonant frequency and should be very low (50-200 Hz). Very small changes on the RX frequency (phase changes) gives different responses for each lock-in amp outputs. All outputs can give a saver discrimination.

BTW, doubling the TX power in differential mode is really nice feature. But a second sound-card is not allways possible. So I will implement both single and differential TX mode driving.
On the other hand, a second sound-card is more flexible (independend volume control, etc.)..

Just a little bit confused now... Anyway, the software should handle all this stuff for me. Its more work to implement.

Aziz

[simonbaker]

Hi Simon,

it is difficult to implement this feature due to fixed RX-resonant circuit (RX LC part). Aziz

Ok, I didn't think it through much. I thought maybe your "extra sharp" resonant coil could be made double resonant by separating the "poles" of the 2nd order equation. Probably too much trouble , I'll think about it on a rainy day...

[gwzd]

I have re-invented a very sensitive capacitive sensor! It detects human. It it is not able to discrimate between man and woman.

Aziz

That's very good Aziz! However, since it has become harder to tell between women and men you should work a bit on that too.
Regards,
R.A.

[Aziz]

New experiment result:

Differential TX feed at 16.2 kHz (doubling TX power, stereo-mode).
TX: 2 x 30 windings, R=2 x 1.5 Ohm, L (total) = 1.69 mH, center tapped coil
RX: 160 windings, ~60 Ohm

1 Euro coin detection distance (air-test):
46 cm (even more )


Aziz

[amtech2005]

Let's make a PC-based metal detector with usb interface !!!

Excellent work Aziz , How long before you are close to some "outside field tests" in the Dirt ??? Very impressive so far !!!.....cheers.........Eugene

[Aziz]

Excellent work Aziz , How long before you are close to some "outside field tests" in the Dirt ??? Very impressive so far !!!.....cheers.........Eugene

Hi Eugene,

I hope I can get into dirt in few month to make some real field testing. Before doing this, I have to build a stand alone search coil in a very stable coil housing and have to please my parents to sponsor me a laptop.

Now I know the requirements for the software and I can start with the coding of the professional version. I bought today a very cheap and low quality sound-card (10 Euro, 16-Bit, 48 kHz) to test the fallback case and also testing the audio signal output (VCO beeping). I will implement the first user programmable MD VCO. This will allow the user to define his own VCO by entering some formulas (mathematical expressions).


Aziz

[Aziz]

Interesting worst-case study:

- 16-Bit very cheap and noisy sound-card
- 48 kHz sampling rate
- output impedance (unknown yet, not adjustable or not found the right settings, but looks like high impedance one)
- search coil operating frequency: 11.9 kHz

Detection depth falls to (air-test):
1 Euro coin detected at max. 16 cm.

The noise density at operating frequency is more than 10-20 times compared to the HD-sound-card. Frequency spectrum is full of spectral noises.

So one only can get the best performance with high quality high definition (HD) sound-cards.

I will measure the output impedance and will lower this when possible.

Aziz

[simonbaker]

Interesting worst-case study:

- 16-Bit very cheap and noisy sound-card
- 48 kHz sampling rate
- output impedance (unknown yet, not adjustable or not found the right settings, but looks like high impedance one)
- search coil operating frequency: 11.9 kHz

Detection depth falls to (air-test):
1 Euro coin detected at max. 16 cm.

The noise density at operating frequency is more than 10-20 times compared to the HD-sound-card. Frequency spectrum is full of spectral noises.

So one only can get the best performance with high quality high definition (HD) sound-cards.

I will measure the output impedance and will lower this when possible.

Aziz

That's interesting.

Are you simulating motion or non-motion detecting?

If you increase the filtering time constant, can you eventually detect the signal, or is the noise "non-stationary" and cannot be reduced by statistical means?

The good news is: you can market your software to hi-fi nuts and audio/video producers to test the quality of their sound cards!

[Aziz]

I have measured different things:

- jitter (total ugly)
- total harmonic distortion (total ugly)
- impedance (ok, 32 Ohms specified and measured much lower for earphone output, 1.1 Vrms output)

This codec-chip (CMI 8738/PCI-SX) is not recommend for such sensitive lock-in amplifier measurements. I have observed clock timing anomalies at 44.1 kHz sampling rate. In this mode, the lock-in amplifier can not be used due to low constant phase shift drift. This might be a chip-related clock generation problem or software driver. On 48 kHz SR, the lock-in amplifier is stable.

Line input is not much sensitive (1.1 Vrms) and the input signal needs to be amplified for more sensitivity (not acceptible).
Microphone input is mono (50..200 mVrms) but can be boosted 20 dB additional. I got more detection distance with mic-input.

This codec-chip is only for your ears - not for your coils. For a professional laptop MD one need really a true HD (high definition) 24 Bit/96 kHz or better sound-card.

Now I will be focusing on software coding (state of the art VCO). I have now a second sound card for audio detecting signalling. And no more plugging again and again TV, MD and so on..

Aziz

PS:
Hi Simonbaker,

I am measuring static signals on the RX coil at the moment (non-motion). When I increase too much the time constants, this will reduce the response time (reducing depth on faster moves). I have not changed the time constants.

[Aziz]

Good news after trivial bad news:

I can run my HD sound-card at 192 kHz sampling rate (output-only). I can achieve up to 187 frames per second for processing. Frame size is 1024 samples.
The output (stimulus) can be accurate sampled. The input is sampled at 96 kHz and converted internally to 192 kHz SR from device driver. This will allow a better SNR and depth sensitivity.

Increasing the sampling rate requires more CPU processing and battery power.


Aziz

[Aziz]

Sound-Cards for Laptop MD.

Creative made the next generation of HD laptop sound-card for Expresscard interface. The former version, the Creative Sound Blaster Audigy 2 ZS Notebook (for PCMCIA interface) works also well (I have this at home ). But the PCMCIA interface is becoming obsolete. All the new laptops have Expresscard interface.

ExpressCard Sound Blaster X-Fi Xtreme Audio Notebook
http://us.creative.com/products/prod...ct=16642&nav=1

It is only a matter of time. I hope the next generation of laptops will imply the HD sound-card on board.

There are small USB sound cards available. But not good enough so far.
Aziz

[simonbaker]

I have measured different things:

- jitter (total ugly)
- total harmonic distortion (total ugly)
- impedance (ok, 32 Ohms specified and measured much lower for earphone output, 1.1 Vrms output)

This codec-chip (CMI 8738/PCI-SX) is not recommend for such sensitive lock-in amplifier measurements. I have observed clock timing anomalies at 44.1 kHz sampling rate. In this mode, the lock-in amplifier can not be used due to low constant phase shift drift. This might be a chip-related clock generation problem or software driver. On 48 kHz SR, the lock-in amplifier is stable.

Line input is not much sensitive (1.1 Vrms) and the input signal needs to be amplified for more sensitivity (not acceptible).
Microphone input is mono (50..200 mVrms) but can be boosted 20 dB additional. I got more detection distance with mic-input.

This codec-chip is only for your ears - not for your coils. For a professional laptop MD one need really a true HD (high definition) 24 Bit/96 kHz or better sound-card.

Now I will be focusing on software coding (state of the art VCO). I have now a second sound card for audio detecting signalling. And no more plugging again and again TV, MD and so on..

Aziz

PS:
Hi Simonbaker,

I am measuring static signals on the RX coil at the moment (non-motion). When I increase too much the time constants, this will reduce the response time (reducing depth on faster moves). I have not changed the time constants.

Yes, not practical, but would be interesting to see if lock-in could cut through noise given enough time, and how long it would take (assuming noise is a stationary process).